You are on page 1of 20

1414 JOURNAL OF THE AOAC (Vol. 60, N o .

6, 1977)

Acid Digestion, Hydride Evolution Atomic Absorption Spectro-


photometric Method for Determining Arsenic and Selenium in
Foods: Collaborative Study. Part I
MILAN IHNAT and HANFORD J. MILLER1
Agriculture Canada, Chemistry and Biology Research Institute, Ottawa, Ontario, Canada KlA 0C6

The hydride evolution atomic absorption spec- Advantages of analytical procedures based on

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


trophotometric (AAS) method for determining hydride evolution atomic absorption spectropho-
As and Se in foods developed and evaluated by tometry (AAS) include nanogram detection lim-
the Food and Drug Administration and Agri- its, a rapid determinative step,2 and capacity for
culture Canada laboratories was subjected to simultaneous determination of both elements.
collaborative study. Twenty-four laboratories
The acid digestion, hydride evolution AAS
provided results for As and 23 provided results
method for determining arsenic and selenium in
for Se levels in 13 different samples consisting
of tuna, swordfish, flounder, oyster, liver, flour, foods developed and evaluated by the Food and
skim milk powder, spinach, kale, and apple con- Drug Administration (FDA) and Agriculture
taining natural levels of As and Se in the ranges Canada laboratories (2) was subjected to a col-
0-15,000 and 0-4000 ng/g, respectively. Ref- laborative study; the results are presented in
erence materials formed a substantial segment this and following reports. A description of the
of samples, and a number of other laboratories study and the overall performance of the
using fluorometry, colorimetry, neutron activa- method are reported here while the following
tion, spark source mass spectrometry, and papers (3, 4) deal with a detailed treatment and
graphite furnace AAS provided confirmative ref- statistical analysis of the results; all reports are
erence values for the remaining samples. A
integral to the assessment of the method.
variety of hydride generation instruments were
used, ranging from commercially available de-
vices to semiautomated and fully automated Collaborative Study
custom-made instruments. Although the accu- Description of Samples
racy of the method was fairly good, between- Thirteen samples representing 6 major classes
laboratory and between-determination (hydride of food (fruit, vegetables, cereal, dairy products,
evolution AAS measurement) precisions were meat, and fish) were selected to evaluate the per-
not favorable. The main advantage of the hy- formance of the method. The samples, sample and
dride AAS method is the rapidity of the deter- analytical solution code numbers, moisture con-
minative step. tents, and weights requested to be taken for analy-
sis are listed in Table 1. They were grouped to
provide each laboratory with 14 samples (12 dif-
The current official methods (1) for deter- ferent samples) for analysis; half the collaborators
mining arsenic and selenium in foods are based received Sample 5A, the other half received Sam-
on arsine generation-light absorption spectro- ple 5B. The shortage of some samples, particularly
photometry (colorimetry) for arsenic (25.006- expensive reference materials, usually required
sending the minimum amounts necessary for the
25.019) and fluorometry for selenium (25.117-
analyses requested. Shortages of orchard leaves
25.120), replacing the Gutzeit method for ar- (Sample 1), tuna (Sample 2), bovine liver (Sam-
senic (25.020) and the thiosulfate volumetric ple 8), and spinach (Sample 9) resulted in either
method for selenium (25.121-25.126), both of randomly excluding some samples or providing
which are now surplus methods. The fluoro- amounts sufficient for only one analysis.
metric method for selenium has a detection limit Apples, skim milk powder, and flour were pur-
of about 10-20 ng, but the arsenic methods lack chased from local retail outlets; flounder and
the capability for estimating arsenic at nano- swordfish samples were obtained from Environ-
gram levels. ment Canada. Reference materials constituted a

1 2
Food and Drug Administration, 900 Madison Ave, Balti- Determinative step and determination refer to the hydride
more, MD 21201. Present address: U.S. Geological Survey, generation AAS measurement step (i.e., digest -> result);
Water Resources Division, Box 25046, Denver Federal Center, analysis and method refer to the entire analytical procedure
Denver, CO 80225. (i.e., sample -> result).
IHNAT & MILLER: HYDRIDE AAS METHOD FOR AS AND SE IN FOOD 1415

Table 1. Description of samples used in collaborative study


Sample Moisture Wt analyzed, Anal.
codes" Description 6 content, %' Ed soln code'

Practice flounder 2.1 _


Pi. P 2 .. •
1 1 NBS Orchard Leaves 6.8 2.000, 2.000 It, 1 2
2 A NBS Tuna 3.0 1.000, 1.000 2i, 22
3 B IAEA oyster 6.7 1.000 3
4 C swordfish-' 2.1 1.000 4
5A JA apple (Macintosh)" 3.8 1.000 5A
5B JB apple (Delicious)' 5.9 1.000 5B

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


6 E flounder 2.1 1.000 6
7 H kale 7.9 2.000 7
8 D NBS Bovine Liver 6.2 2.000, 2.000 8|, 82
9 K NBS Spinach 4.6 2.000, 2.000 9|, 92
10 F IAEA flour 10.8 2.000 10
11 L skim milk powder' 4.9 2.000 11
12 G flour' 9.6 2.000 12
13 B IAEA oyster 6.7 1.000 13
14 H kale 7.9 2.000 14

Standard Solution Ampoules

A As and Se A,, A 2
B As and Se Bi, B2
C As and Se C,, C2
D As* —

" First column is collaborative laboratory code; second column is reference laboratory code.
6
All samples were dry or freeze-dried materials. The 2 flounder samples, 2 oyster samples, and 2 kale samples
were blind duplicates.
c
All samples except 1, 7, and 14 were dried under vacuum 15 hr at 70°C; Samples 1, 7, and 14 were dried under
vacuum 24 hr at 90°C. Mean of 2-5 analyses.
* Weight requested; actual weights used differed in some cases.
• Subscripts 1 and 2 for samples and ampoule C refer to first and second digestions, respectively. Ai, B|, A2,
and B2 refer to undigested solutions prepared from ampoule contents.
1
The swordfish, skim milk powder, and flour samples were identical to those used in a previous collaborative
study (5) where they were numbered 8, 2, and 9, respectively.
" One half of the collaborators received Sample 5A, the other half received Sample 5B.
h
Used for spiking Samples 5A and 5B.

substantial segment of samples and included NBS and Se in the ranges of 0-15,000 and 0-4000 ng/g,
Standard Reference Materials Spinach (SRM respectively. Only the apple samples (Samples 5A
1570), Orchard Leaves (SRM 1571), Bovine Liver and 5B) were fortified with As by collaborators,
(SRM 1577) and Tuna (not yet issued); Inter- because native levels were low and we wanted to
national Atomic Energy Agency (IAEA) oyster evaluate the performance of the method for this
(MA-M-1) and flour (V-2/1 (1974)); and kale food matrix.
(H. J. M. Bowen). In addition, 3 samples, sword- Four glass ampoules containing standard solu-
fish (Sample 4), skim milk powder (Sample 11), tions of inorganic As (III) and Se (IV) in IN
and flour (Sample 12) were used in the 1973 col- H2SO4 at concentrations of 625 ng As + 625 ng
laborative study of the fluorometric method for Se/ml (ampoule A), 1250 ng A s + 1250 ng Se/ml
Se (5); extensive analytical data concerning Se (ampoules B and C), and 1250 ng As/ml (ampoule
were available for these samples. D) were included to monitor instrument perform-
For ease of handling, all samples were shipped ance and the digestion step (ampoules A, B, and
in screw-capped glass vials as either dry or freeze- C), and to provide a solution of As for fortifying
dried materials. Flounder, swordfish, and apple Samples 5A and 5B (ampoule D).
samples were prepared by homogenizing in a
blender, freeze-drying, grinding, and mixing; the Description of Study
skim milk powder was ground to a fine powder; Each collaborating laboratory was supplied with
the NBS and IAEA samples and kale were used the 14 samples (unidentified), and 1 practice sam-
as received. During packaging, 2-5 vials of each ple (flounder containing 15,000 ng As and 1500 ng
sample were set aside for determination of mois- Se/g). Collaborators supplied their own reagents
ture; it ranged from 2.1 to 10.8% in the different and equipment, and were instructed to request
samples. Samples contained levels of native As elemental Se for preparing standard solutions if
1416 JOURNAL OF THE AOAC (Vol. 60, No. 6, 1977)

that reagent was unavailable. A copy of the method provide a possible third pair of blind duplicates
(2), additional instructions, and report forms were should practice results be reported and be accept-
included. able. Ampoule solutions A, B, and C were pre-
The study was designed to amass a large body pared to contain equal concentrations of As and Se
of information in an attempt to relate the per- so that, when diluted, concentrations would be Al5
formance of the sample treatment and determina- 10.0 ng/ml; A2, 1.0 ng/ml; B ^ Ci, and C 2 , 20.0
tive steps to a number of variables. Parameters ng/ml; and B 2 , 2.0 ng/ml. Incorporation of these
considered were analyst experience, type of hy- standard solutions was expected to provide infor-
dride generation apparatus and technique, repro- mation on detectivity, accuracy, and systematic
ducibility and type of calibration curves, instru- error of the determinative step, and effect of

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


mental baseline noise, analytical precision and digestion on analytes in simple matrices.
accuracy, detection limits and sensitivities of the Collaborators were requested to follow specific
determinative step and the method, absorbance- sample digestion and determination sequences.
time peak shape and reproducibility, digestion Laboratories were randomly assigned 1 of the 8
technique, digestion and determination sequences, permutations of sequences. This approach was
recovery of inorganic As and Se from untreated adopted in an attempt to isolate any systematic
standard solutions and digested standard solutions error due to differences in performance of collabo-
and fortified samples, undigested and digested rators as they became more familiar with the pro-
reagent blank magnitudes, variabilities and origins,
cedure. Samples were grouped according to analyte
sample type, and analyte levels. Consequently, a
levels to facilitate AAS measurements; for those
large number of analyses/determinations were re-
samples with moderate to high concentrations, an
quested from each laboratory, together with de-
instrumental scale expansion of Xl was recom-
tailed information regarding the experimental
technique. The raw experimental data in the form mended; a higher scale expansion was suggested
of recorder charts were also requested. Many labo- for low level samples. The magnitude of the higher
ratories were invited so that a number of different scale expansion was left to the discretion of the
commercial, modified, and custom-built hydride analyst because it depends on the performance of
generators would be included and sufficient num- individual instrumentation. It was requested that
bers of collaborators would use the same types to all determinations be conducted in duplicate to
permit statistical assessment of several specific enable evaluation of this step of the method.
types of generators. Another, smaller group of The greatest difficulty in the design of the study
laboratories, denoted reference laboratories, was was related to the wide range of hydride generators
approached to provide independent analyses of the in use with divergent detectivities, solution volume
samples, particularly the non-certified samples, requirements (1-100 ml), and other performance
using analytical techniques other than hydride evo- characteristics (e.g., standard curve linearity), and
lution AAS. It was expected that these results the desire to make the experiment suitable to all
together with certified reference and other data types of generators used by collaborators. As a
would give good estimates of As and Se concen- compromise, the digests of 1 g high analyte level
trations for evaluating the hydride method ac- samples or 2 g low level samples were diluted to
curacy. Manufacturers of hydride generation in- 100 ml, and collaborators were requested to use
strumentation and AAS equipment were con- aliquots suitable to their particular instrumenta-
tacted for relevant information on operating de- tion, diluting if necessary to get absorbance read-
tails and performance characteristics of, in particu- ings on scale.
lar, the former, so that this information could be Collaborators were instructed that the digestion
on hand when collaborative results were received. procedure was firmly specified and should be fol-
lowed closely. Other details of the method, how-
The scheme of sample coding and the analyses ever, particularly those related to the hydride gen-
requested gave 6 pairs of duplicate results, and 7 erator, the acid and reductant concentrations, and
unpaired sets of data. Samples 3 and 13 as well the measurement step, were left to the laboratories
as Samples 7 and 14 were, unknown to the col- to decide and implement appropriate procedures
laborators, identical samples of IAEA oyster and for their particular instrumentation. Samples
Bowen's kale, respectively, serving as blind dupli- were assembled in 4 groups of 6 so collaborators
cates and augmenting the 4 known pairs (lj, 12), could use one or more 6-element digestors, 12-
(2lF 22), (8j, 82), and (9X, 9 2 ). In addition Sample element digestors, or whatever their capability al-
6 was identical to the practice sample which could
lowed. They were instructed to use the practice
sample provided to gain experience with the
method and to ensure that everything was func-
Contribution No. 961 from the Chemistry and Biology
Research Institute. tioning normally. All samples were dry powders,
IHNAT & MILLER: HYDRIDE AAS METHOD FOR AS AND SE IN FOOD 1417

and were not to be further dried or exposed longer ment to determine concentrations, and one for
than necessary to complete the weighing. each element (at a high recorder chart-speed, at
The ampoules contained standards in acid solu- least 150-200 mm/min) to provide information on
tion. Working solutions were to be prepared as peak shape. Fit breaks in the determination at the
follows: Pipet 4.00 ml solution from ampoule A end of a run. A long break, particularly if flame is
into 250 ml volumetric flask and dilute to volume extinguished, may warrant a recheck of instrument
with the 5% H 2 SO 4 + 30% HC1 diluting solution, parameters. If more than one day is necessary to
or similar solution geared to hydride apparatus to complete the determination, carry out one-half of
give working solution Ax. Dilute 25 ml solution Ax the determinations on each of 2 days. Leave chart
to 250 ml with the same acid mixture to give running continuously during all the determina-

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


solution A2. In like manner, prepare solutions Bj tions, except during high speed determinations, to
and B 2 from ampoule B. Pipet 4.00 ml of solution provide information on noise, drift, etc., under con-
from ampoule C into 25 ml volumetric flask and stant instrument settings. For high chart speed
dilute to volume with water to give solution C; runs, record only the peaks, but include sufficient
duplicate 10 ml digests diluted to 100 ml are de- baseline before and after the peak.
noted solutions Ci and C2. Prepare solution D by Collaborators were requested to submit the
similarly diluting 4.00 ml of the contents of am- original high chart speed tracings, and, if possible,
poule D to 25 ml with water. Add 5 ml solution D the entire original chart record with appropriate
to 1.00 g Sample 5 in Kjeldahl flask to yield forti- notations of parameters and analysis conditions.
fied sample. Collaborators were requested to report absorbances
Rectify charred samples as indicated, salvaging of the undigested reagent blanks and the standard
sample if possible. Report results for both the solutions for all analyses and to indicate whether
charred sample and a repeat sample if possible. an actual absorbance or other instrument mode
Pipet 10.0 ml solution C into Kjeldahl flask, add was used. Concentrations of As and Se (or at least
digestion acid mixture, and proceed with digestion. absorbances, if concentrations were not calculated),
Repeat process with second 10.0 ml aliquot. For found in the 4 digested reagent blanks RB^ RB 2 ,
Sample 5A or 5B, weigh 1.00 g sample into Kjel- RB 3 , and RB 4 and volumes used for hydride gen-
dahl flask, pipet in 5.00 ml solution D, add acid eration were requested. The nature and use of the
digestion mixture, and proceed with digestion. 2 types of blanks were explained diagrammatically.
Ensure that each standard (calibration) curve For all samples and ampoule solutions it was re-
contains 5 points, including reagent blank, cover- quested that aliquots of solution taken for hydride
ing range of instrumentation at each scale expan- generation, analyte concentrations, and contents
sion. Prepare all standard solutions from Se° and in the aliquots be reported; in addition, for sam-
As2O3 (commercial standard solutions may be ples, analyte concentrations (ng/g) were requested.
used). Standard solutions and associated reagent Information on the following instrumental and
blanks are not digested; all samples, solution C, flame parameters was requested: atomic absorp-
and associated reagent blanks are digested. tion spectrophotometer and recorder makes and
Use whatever solution volume is required by models, resonance line wavelength, monochromator
generator, taking up to 20 ml aliquot from sample slit width and spectral bandpass, scale expansion,
solution and making any dilutions to required vol- signal damping, lamp type and operating current
ume with H2SO4-HC1 mixture. If NaBH 4 and or wattage, recorder input voltage for full scale
H2SO.|-HC1 concentrations specified are unsuitable, deflection, chart speed for routine and high speed
use concentrations appropriate to specific situa- runs, flame gases and flow rates, burner head type
tion. Try to use same acid matrices for both As and slot dimensions, and height of center of light
and Se, and adjust acid content with HC1 because beam above burner. Inquiries were also made re-
5 ml H 2 SO 4 remains after digestion procedure garding the use of draft shielding for the flame,
which is firmly specified. whether spectrometer parameters were optimized
When samples give off-scale absorbance readings, using criteria of sensitivities (highest signal) or
repeat analyses, using <%th sample solution detection limits (highest signal/noise), and whether
volume. It may be necessary to dilute solutions standard solutions or hydride generation was used
for automated equipment. for optimization. The following information con-
In determination sequence (hydride generation cerning the hydride generation apparatus was so-
and measurement), use the standard solution and licited: make and model of the generator, and
reagent blank absorbance readings preceding the whether commercial, modified, or custom-made;
sample readings in each run for calculating As volume of solution required by the generator and
and Se concentrations for that run only. The entire actual volume used in this study; acid mixture
determination consists of 18 runs, 8 for each ele- used in this study, whether it differed from the
1418 JOURNAL OF THE AOAC (Vol. 60, No. 6, 1977)

collaborator's normal operations, and whether the unsatisfactory to themselves and hence reported
concentration was optimum for both arsenic and no results. In addition, 11 laboratories reported
selenium determinations; nature and form of re- reference analyses for methods other than hy-
ductant (NaBH 4 was specified in method); re- dride generation.
ductant solution concentration and composition
All hydride generation AAS results for sam-
(e.g., NaOH stabilizer concentration); amount or
volume of reductant used per determination; and ples are presented for arsenic and selenium in
type of auxiliary carrier gas and flow rate. Col- Tables 2 and 3, respectively. Although most col-
laborators were asked to describe operating in- laborators conducted duplicate determinations as
structions and construction details if they differed requested, some reported only averages and not

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


from those for commercial apparatus or if custom the individual results. The submission of raw ex-
designs were used. It was emphasized that volumes perimental data, however, made it possible for
and operating conditions optimum for specific ap- the authors to extract these individual data for
paratus used should take precedence over instruc- some collaborative sets to provide a larger data
tions provided by the initiating laboratories. Other base for calculation of within-digestion variance.
information requested dealt with sample digestion In addition, concentrations could be calculated
details (charring, foaming, etc.), and sources and
purities of NaBH 4 and other reagents. Collabo-
in several other instances where results were re-
rators were also asked to indicate the experience ported as "not detectable," "less than . . . ," or
of the participating analysts with the wet diges- "0." These results calculated by the Associate
tion and hydride generation. Referees, are included with reported data in
Reference laboratories were free to use analyti- Tables 2 and 3. When only one of the determi-
cal methods with which they had had experience nation pair warranted recalculation, levels were
and in which they had confidence. They were recalculated for both determinations to obviate
asked to conduct any number of analyses/deter- artificial inflation of standard deviation due to
minations for As and Se on any number of samples incursion of a measurement/calculation bias
provided, and to submit only the data with which likely to be different from that of the collabo-
they were satisfied. Sample identities were not rator. Calculations were based on raw experi-
revealed, but ranges of As and Se present (0-0.1,
mental data provided in the form of recorder
0.1-1, 1-10 jtig/g) were given to facilitate analysis.
Samples were coded differently than for collabo- charts, reported peak height measurements, or
rative laboratories in an attempt to preserve sam- reported weights or concentrations of arsenic or
ple anonymity in instances where the laboratory selenium in the solutions and do not necessarily
was requested to serve the dual role of collabo- reflect complete remeasurement/recalculation;
rator and reference laboratory. Reference labora- calibration graphs provided by some collabora-
tories were requested to report mean As and Se tors were used when deemed suitable. Calculated
levels found, standard errors or confidence limits, concentrations are corrected for digested reagent
and number of analyses/determinations conducted. blanks in the same determination run as the
They were also asked to report on the analytical sample, regardless of whether the blank was di-
method used, together with literature references
gested with the sample. Negative values are in-
and details and to mention any problems en-
countered.
cluded for the purpose of estimating precision
and systematic error. More digits than war-
Results and Discussion ranted by the precision of the method have been
Of the 55 laboratories that agreed to partici- retained. Each datum is for one determination
pate in this study either as collaborators or ref- unless otherwise indicated; data include results
erence laboratories, 24 submitted collaborative extracted, by the Associate Referees, from high
analytical results on arsenic and 23 submitted recorder speed runs and other raw data not used
results on selenium. One collaborator who did by collaborators. The body of sample concentra-
not have HC104 digestion facilities, analyzed tion data used for statistical analysis is indi-
only the undigested ampoule solutions. Two cated by underscoring.
others used Zn or Al as reductants rather than The sample concentration data were scanned
NaBH4; their results are presented separately for extreme values, and a number of results were
with data from other laboratories which used rejected subjectively prior to statistical calcula-
different sample treatment steps (3). Four labo- tions for the following reasons: All results for
ratories attempted analyses but obtained results both elements from Collaborators 3 and 34 were
Table 2. Collaborative sample results (ng As/g) for the hydride AAS determination of arsenic in foods9
Sample/solution
g
Coll. 2 2 3 4 5A 5B 6 7 8 8 10 n 12 13 14
h h 1 2 i 2 '1 '2
g
t 4(> b 157b
ILL

1 8745 10270^ 2897 3000 11385 3199 1000 - 14080 117 63 150 ND^ ND ND 11180 115
9120 11220^ 2760 3600 12020 2960 880 13980 112 55 38* 162 165^ ND ND ND 12520 115
lp. 12?- 4?-
-p. 4?

50 100 300 <100 120 <100 <50 50 50 <S0 50 <S0 300 50


t
4 9280 9910 1000& 1590 11100 1760 2290& 7830 99 20 24 110 104 <2Ct- <2cr 11260 69
9280 9730 100CT 1460 11100 1710 22OO& 7610 99 20 *2(F- 110 104 28? 29? 7- 9890 69
26? -26?
62?
h h
5 8371h 2773^ 2S94-- 11250^ 2919- 998~ 14881^ 100^- SO* s*e ND| 11239*- 92*
r: HYDRIDE AAS METH

8800 s 8371? 2623? 2614? 11400? 2959? 968-? 14087?- 98? 49? 134? isT? 12s 11239? 92?
8900^ 8596H 2723? 2594? 11650^ 2979^ 1028^ 16667.E. 95? 38? 44? 121? 151- 7? -Z~ T? 10989 s 90?
8450- 28352
O
8b 1605 s 3720 12900 3840 790 14200 40 145 160 150fi 190 11580 65

10 9545 2942 2875 14160 3197 1100 16420 42 30 125 0 12540 65


9245 3115 11040 3047 1020 16500 85 35 168 0 0 0 13720 107
-2| -6? 2-?
O?
o
F
11 106374 3204 10397 1171 16284 236^ 252* 453* 2251 167 11740 171
4677-1 2313 11113 600 1855^ 12712J- 476-L 231& 403^ 465s 117-1 205 11718

12 9540 3128 _ 10509 3643 1240 15684 39 16 146 ND ND ND 11758 251£ 8


8116 2788 10384 3214 1050 14452 36 6 71 ND ND ND 11591 246s
-35? £
-15? 1

13 10000 9350 3230 3000 8660 2670 720 12500 103 40 140 ND ND ND 8000
10250 8750 2270 2000 9500 2600 780 12500 105 38 128 ND ND ND 8500
4?
-i"i? -a?
h h
15* 900CT 10900^ 409CT- 3560*- 12070**- 2970^ 700^ NDe - ND ND ND
7070 s 10160- 4280? 3875- 10730-3 668? 14260^ -38- 33? _29O?>£ -78 s -20? 11910? 100?
9500^ 9780S 3060^ 4010S 11100^ 2790? 2 1462O£ 13870?
2550?
2970(4 2 )-

16 7500- - ND 1800^ ND - -
35(7
i >J -
9500? 750^'* 120^? 285? s. 80^ 100? 218?., -' 168JL •
9500? I 1130? 2630^'^ 655?^a Si 155? 118 i—*

(Contined) CO
Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020
Table 2. (Continued)
Sample/solution O
Coll. l 2 3 4 5A 5B 6 7 8 8 9 10 ll 12 13 14
h 2 h 2 1 2 '1 2

18* 7700 9100 400s 450| 4200s 600 1060 4700 70 30^ 40 80 90 -40l . -20 5700i 75
8300 7700 600f 60of& 470CT 500 1060 4400 60 30=!- 35 150 120 -40l -20 570C4- 80
8300 9200 7002 850 4600i 900 1100 4500 65 250i 50 110 90 lOOi 50 lOOOoi 65
500i 5200s 680 1100 6560 80 2001 100 110 80 10? 50 97001 70
1070
19 9200 2900 2900 12100 1020 320 0 100 18 130 9 0 0 12900 35
10700 2600 3400 13100 970 370 100 g125 20 160 5 0 3 13900 55
95
f
22 8830*'°. 115002.'° 2960*- 13900*- 102*- 11* 16*^2.116002. 101*
86485 27065 116065 28465 9365 141725 1155 505 1265 255 55 185 121615 95S.
90505 27965 117105 30555 10165 135735 1105 455 1215 205 105 235 120605 1055
23h 9100 8900 4300 4400 12900 4200 2650S. 18900 12600
24* 6700*- - 1980*- 255C& 7102- 109002. - 502- - 85802-
6100£>t 2900i>A 26755 lO6oS 3200^ 530^' 10725- 68s 85 s 25s
s s 1 8675I
53755>i 1050^'i 2525 s 8200 2075s'! 895 s '•i 11175 s 63^ 8 3"? -25- -3- 98^''•"- 850CT 120&J
7300 s 7975s
28 8950 8350 f,a
3610 5060s- 1210 1015ol 195 125. 185 170 85 150 13300 140
8850 3530 470OS- 1270 6150^-'-^- 63001 185 <50— 110 <100- <50^- 60 10800 135

29^ 162004- 6610 s 74304 4310 1350(41 68904 4Ol,s 460 s 4610 s 790s 950s- 0 660s- 0 5 no 1 853O1 35l
1040CH- 3210 s 3610 1 4210 849O 571O 1 300 s 7550s 470s 900s 820s- 240l 158001

30 328C& 12130*. 3010^ - iwsS- 1S42- ND 23825^-


901 if 3760 s 10026s- e U
2943s- 5? 1 32l 164- 10221- 8^
s
87 srf
10213- 2522 s 11623 s 1103 - "63 87^- 137- 11825 s 96 s
JOURNAL OF 1

9412- 12169 s 3020^ 1065^ 12597-'1 12606 s


9440s a
31 11760 *, - - - 20802- 23580- - - - - - - - - 32SS()Ji -
E AOAC

11080^ 2720 s 13720 s 3480 s 24505 16420S. 11300^


12630s'-! 3280 s 11700s- 3080s 2180s 15870- 1205Or • " • ^

1285s'-1- 11830- 14100-

34- 350 3500 2200 2800 400 800 350 900 250 700 350 700 1800 350

39 9708 9287 s
7146 2719 12675 3074 - 1337 14950 101 42 66 130
9067 6253^ 117* ND ND 7 12529 107
2429 12065 3127 1139 14227 102 61 53 144 146- ND 21 12137 107
Ji- -12"
1. 60, No. 6,1977

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


h h 14900h . 3700h 35300 <300 <300 <S00 <300 i3£00 <300 «
41 oUU
1145Og 10950S-4 1470O5'-1- 3520 s 197006 _65£'-£ -35£ -105-'•= -8Osfl270Og -lOO^f §
s s 6 s s
s 8
10500- 1125d£'J- 157OO£'4- 356QS 19200 -60 "- -.50 s -180 £ > - a -35 540 '* • ^^ -160 - >
10150- 79OO£'l 9700 s '- 1 - 12000s ^

10960 20 340 1370 243 158 238 148 3 63 170 160


12380 0 280 1700 238 163 213 108 10 53 550 138

" All data are on an "as received" sample basis. Designations of data, by way of (example) are as follows: reported by collaborator (1000); calculated by Associate
Referees (1000°); reported by collaborator and considered for statistical analysis (1000); calculated by Associate Referees and considered for statistical analysis (1000s);
reported by collaborator and used for calculations (1000);calculated by Associate Referees and used for calculations (1000f)- For reported data, calculations were checked
using the ng, ng/ml, solution volumes, and sample weights reported, to correct some obvious arithmetic errors, and data have been rounded-off to, at best, 1 ng/g; a
an extra digit was extracted by the Associate Referees in some cases of excessive rounding-off. The number of digits is unrelated to precision. Each result reflects 2
one determination unless otherwise indicated. Appropriate reagent blank corrections were not always made by collaborators. D
6
Shortages of some samples resulted in insufficient a mounts being sent to some laboratories for duplicate analyses requested. This laboratory ascertained by visual H
inspection and consultation with one of the Associate Referees that Samples 1, 8, and 9 were NBS Standard Reference Materials. These samples were available in 02
Laboratory 1 and were used to perform duplicates I2, 82, and 92.
" Not determined or not calculated.
d
Reported as not detectable. 3
" Results were calculated by the Associate Referees; see text.
1
All results from Laboratories 3, 34, and 49 were excluded from statistical calculations because most results from these laboratories appeared to have large system-
atic deviations; in addition, 5 results of Laboratory 3 are in the "less than . . ." form and unusable. Other "less than . . ." data from other laboratories were also
excluded. The 5 results from Collaborators 28, 29, and 41 for Sample 5B were excluded because of erroneous sample handling
" Results excluded from statistical analysis as outliers because the laboratory mean for the sample deviates excessively from the mean over all laboratories.
* Mean of 2 determinations.
1
The 42 ng/g reported appeared in error; recalculation gave 118 ng/g based on the 2.36 ng/ml reported.
' Results excluded from statistical analysis as outliers because within-digestion variance appeared too large relative to that for the other laboratories.
* Results were rounded-off to 2 figures by Collaborator 15 who reports "we have reported our values well beyond their significance." An additional figure was obtained
by the Associate Referee from the reported ng/ml; values have been rounded-off to 3 figures or 5 ng/g.
' Value of 6030 ng/g reported by Collaborator 15 seems in error due to erroneous value for volume used in calculation: 12070 ng/g was recalculated by Associate
Referee from the 1216 and 599 ng reported and 10 and 5 ml volumes, respectively, obtained from recorder charts.
™ Second digestion of Sample 4.
" Calculated by Associate Referee from ng/ml reported on recorder charts. Data obtained at both low and high scale expansion are listed although some data at low §
scale expansion are negative and inaccurate.
0
Calculated by Associate Referee from reported ng/ml and sample weights; differ slightly from reported values of 8800, 2700, 11100, and 14 ng/g, respectively.
p
Collaborator24 remarked that high reagent blanksand irreproducible calibration curves led to inconsistent and inaccurate results and suggested that these results
for arsenic not be considered.
" 5 ml of solution in ampoule D added instead of 5 ml of 6.25-fold diluted solution as instructed.
r
Average standard curves were used because a sufficient number of standards were not always included in each run.
8
Solution D was not added as instructed.
' Laboratory 39 knew identity of Sample 9 and used NBS Spinach for determination 92.
" 4 ml of solution in ampoule D added instead of diluted solution as instructed.

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


Table 3. Collaborative sample results (ng Se/g) for the hydride AAS determination of selenium in foods*
Sample/Solution

58 57b 4115 3894 1868 2700 ND^ -£- U98 107 1000 1034*" 8 292 105 684 1946 125
73 2i 4195 3595 2067 2839 ND 1398 102 1150 1064^ 25 263 95 764 1936 117
-245

3- <50O - 2000 - 990 1100


4 3470 3620 1700 2530 1320 li 938 890 J>4 194 125 563 1650 195
J64(lj)2 3510 893 54 T68 565 1650

las
3530 1590 2470 55J 1120 131 890 125 139

5 ND 187*1 1696^ S00-? ND e ND ND 20\


-11s 48s 135-i 195- -12O|>£
-T55 IP 1755--8 'iMeS.t Jss 171- -330^'^ _4755.£ -Iio5.£ '•US'* -llO^'* 3?
s!> 310 - 5160 - 1820 3170 270 1070 915 60 5J) 525 2020 245

10 22 - 37U 3800 1900 2777 0 1335 147 962 68 265 40 687 1965 170
4021 3710 1920 2777 "1460 127 935 180 440 57_ 725 1900 80

11 190 - 4176 - 1540 2946^ M 1149 202 973 829 223| 379i 38 737 973 128
278 3597 1413 12541 1532^ 202 544 725 344* II7J- 54 658 1139

12 86 3548 832 2443 1159 180 491 132 80 703 590 121
- - 21
178 3760 687 2502 1149 193 539 173 123 672 720 166
32
13 ND 3200 2650 1800 2400 120 950 45 200 100 1600 - =-1
40 2900 TWO" 2400 110 7B" "67~5 1500
-3s
I -35 §
jp.
675 2100 175 3!
14 125 3600 3650 1950 2750 50 1500 200 1000 25 230 100
150 - 3600 3450 1850 2700 0 1270 150 900 10 140 125 575 2150 225
2800 1300 775 tr1
2350 1200 850

15* ND soi 39M& 203 Oil 2950- iooh 796!! ieit£ 3


373O| 30055 16255 n-h •JI
-375Ij 1365^ 405'i 10235 5955 -1085'i 175 s '-'- 6835 U2<fis -63s
3325^ 3945- 205Q5 28755 17755 1685.1 97OE 38S-S 555.1 778$ 1940
t -155.1
-i 2475^ I
3658(42)

16 - 4300^ - 800* ND 3600i ssoi ND 900± 1700^ 200^


51205 690" 1815| -3705-1 158Q5 5035 9481 3555 225| 1445 s 183|
55s 42805 10055 2530- -205.1 "1570s 8485 25|5 850s 1455 s 1755
65Q5

247^ 357-} 3400 370O* 300 2000 94 124 158 0 97 167 - 594 500 119
307i 2871 75BB 3400l 355 2200 79 179 154 87 87 257 559 300 119
400i 30(4 M 0 0 4800l 400 2300 200J 84 150 200 7 27 300 650 500 104
3001 30(H 150 2100 3 154 150 150 117 57 300 "550 $2-
42 °?J
1070^- 540 - - 290*s1 2200^ 131-
22895 9025 11745 05 10785 1405 5455 255 605 705 285 1236f
B4&5 8705 12145 05 U085 5605 30 s 605 55E 3155

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


3730S- 21302 2400R 12S0± 12S± lllSfi-s
635 40385 37585 2289 s 24535 195 1345 s 120 s 1178 30* 202^ 774~ 2065^ 97^
75S 4135 s 3625 s 1971 s 23085 3T-5 11055 T265 10435 3l« 1827i
38285 21745 -75 1215 I? 2062- 99 s
-3
2130 - 2030 2240 <100 1900 150 445 250 70 720 1680 140
250 1930 1970 2430 130 1260 TfO 620 <S0 75 665 1950 95
1055 2105 405 in
2305 -705 T5
ND 6100^ 7340^ 4539^- J849^ 164(& 24& 200h 270^-
- t
-125 49655.1 29505.1 4045 1950s 2080^'^ 1325^*1 287^'^ 2160s* fsS-4 H
725054 28055.4 53805.1 2375 JTJF 21605-4 200O5-1 2695-K •5625.1 28054 150054 3O7O S ' £ 365s*
5845 s ' 1 r?
25355.4 w
4580^ - SiflS eotfi 0s a
s s
58105 900 s 36005 0 990 •OS
5060= ?105 3350 s 4305 7755 54§5 S
4610 s "736s
a
2600 2920 880 2200 570 320 ND 640 1040 ND
2480 2920 760 420 320 520 , 92
1240(62)fl 27/5?/U22)A'
m °
1000(6 2)~'B 950(122)i.

160 25 25 1280 25

39 126 118 3872 3667 1894 2067 ND 186 142 1055 1030 116 83 686 2077 152 §
148 121 3583 3821 2023 2488 56 1447 162 T255 1136 IS 754 2003 107
3
41 4400^ - 29<Wg UOCfz JSSOf 1200% 300\ 700^ 10&
655 455 38505 208 OS 2730- 150- User 495 s ' 1 1570- 1380- lO^ 300 s 75^ 620 s 212Of 115 s
955 905 44705 21505 29005 T405 J46QS 590£'£ 14455 10305 1605.1 |55 735- 1360- T355

48 78 - 3425 - 1910 2605 27901 1445 2931 235 90 613 2190 148
a
35 4045 2045 2670 26951 1385 1181 145 103 605 1935 173 o
S!
»-* See Table 2.
' All results from Laboratories 3 and 34 were excluded from statistical calculation as most results from these laboratories appeared to have large systematic devia-
tions; in addition, 8 results of Laboratory 3 are in the "less than . . ." form and unusable. Additional data reported on duplicate digests 42, 62, and 122 were not con- 8
sidered.
™ Second or third digestions as indicated in parentheses.
" The 1240 ng/g reported by Collaborator 11 appeared in error; 1532 ng/g was recalculated from the 15.32 ng/ml reported.
' Calculated from ng/ml reported on recorder charts. Data obtained at both lowand high scale expansion are listed although some data at low scale expansion are
expected to be inaccurate.
» Some data as reported (4255 ng/g for 2i and 501 ng/g for 12) differed substantially from the check calculations recorded here.

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


1424 JOURNAL OF THE AOAC (Vol. 60, N o . 6, 1977)

excluded because most results from these labora- atory, and sx is the precision expected when anal-
tories appeared to have large systematic devia- yses are performed by different laboratories. For
tions; in addition, 5 results for arsenic and 8 the 6 samples analyzed in duplicate, it was pos-
results for selenium reported by Collaborator 3 sible to arrive at estimates of all 5 standard
were in the unusable "less than . . ." form. All deviations. Analytical data for the 7 samples
arsenic results from Collaborator 49 were ex- analyzed once by each collaborator furnished
cluded on the basis of large systematic error. A only 2 of the standard deviations, sD and sx, and
number of other results were also excluded be- the sum, V * L 2 + SA2- In the 12 cases (6 for
cause within-digestion variance appeared too each element) for which sh, sA, and sD are avail-

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


large relative to that for the other laboratories, able, usually sL > sD > sA. This observation
or the laboratory mean for the sample appeared points to some imprecision in the determination
to deviate excessively from the mean over all step, most likely arising from the hydride gen-
laboratories. (A subsequent paper in this series eration-transport phases rather than the actual
(4) will give a more detailed discussion.) atomic absorption measurement. The between-
The remaining sample data were subjected to digestion variance sA2 was the lowest of all 3
analysis of variance to arrive at estimates of variances, indicating that the procedure of sample
the between-laboratory standard deviation, sL, digestion contributes little variation to the final
between-digestion standard deviation, sA, and result. In fact, sA2 was negative at times (and
within-digestion (between-determination) stand- sA imaginary), suggesting that its real value was
ard deviation, sD. Estimates of the precision a very small positive number. The largest stand-
standard deviation, sR, and the overall or repro- ard deviation, sL shows the existence of labora-
ducibility standard deviation, sx, were obtained tory-dependent systematic error in the method.
from the relations sR2 = sA2 + sD2, and Sx2 = Coefficients of variation (CV) listed in Tables
SL2 + SA2 + SD2- Standard errors, sx, of the mean 6 and 7 also reflect the performance of the
were computed from sx2 = [sL2 + sA2/nA + s D 2 / tested method. The CV values based on sx are
wAnD]/nL, where nL is the number of labora- fairly large for both elements and all samples,
tories contributing to the mean, nA is the num- indicating a wide and unacceptable range of sys-
ber of independent analyses carried out on the tematic errors in the method. Coefficients based
sample by each laboratory, and %> is the num- on sR are smaller in comparison, but still seem
ber of determinations on the digest. In this too high, even for samples containing /xg/g
study, the values nA = 1 or 2, and wD = 2 were amounts of analyte.
taken to represent the data fairly; the fact that We attempted to establish the accuracy3 of
nA and %> were not consistent for all labora- the method by comparing the analytical data
tories would have only a small effect on the with levels for the samples obtained by inde-
estimation of sx2 and was ignored. Statistical pendent analyses using methods other than hy-
results are summarized in Tables 4-7. dride generation AAS. Tables 8 and 9 detail
Tables 4 and 5 give mean concentrations of results from reference analyses of the collabora-
arsenic and selenium found in the 13 samples, tive samples, together with other data available
together with the corresponding standard devia- from the literature and other sources. A total of
tions. Analyte concentrations used in most cal- 6 different analytical methods (4 for arsenic and
culations were not corrected for moisture con- 3 for selenium), instrumental and destructive
tent, the only exception being the data used for neutron activation, isotope dilution spark source
accuracy comparisons in Tables 8-11. The inter- mass spectrometry, fluorometry, light absorption
pretation of each of the 5 standard deviations photometry, and graphite furnace AAS, were
shown is as follows: sL represents systematic used to measure concentrations of arsenic and
error among laboratories, sA reflects errors aris- selenium. The lack of biological standard refer-
ing from the sample digestion procedure, sD is ence materials (only 2 NBS SRMs, Orchard
the precision of the determination step, sR is the Leaves and Bovine Liver, are at present certi-
precision of analysis expected in any one labor- fied for selenium, and only 2, Orchard Leaves
3
This report of the Associate Referees was presented at the Accuracy in this work is taken to mean the extent of
90th Annual Meeting of the AOAC, Oct. 18-21, 1976, at agreement of the method mean over all laboratories with a
Washington, DC. reference value.
IHNAT & MILLER: HYDRIDE AAS METHOD FOR AS AND SE IN FOOD 1425

Table 4. Statistical evaluation of collaborative results for arsenic (ng/g)

Between-lab. Between-dig. Within-dig. Precision Reprod.


Sample Av. concn s t d dev ., s t d dev., std dev., s t d dev., s t d dev.,
(code No.) found" SL SA SD SH Sx

Skim milk powder (11) 0.12 (14) 24.9* 10.4 27.0


IAEA flour (10) 31.2 (14) 49.7h 32.0 59.1
Flour (12) 31.4 (15) 60.2* 26.6 b
65.8
NBS Bovine Liver (8) 43.3 (17) 25.9 22.8 17.6 27.0 37.4
Kale (7/14) 93.1 (17) 26.2 23.4 11.4 26.0 37.0
NBS Spinach (9) 114.0 (16) 45.5 c 23.1 20.3 49.8
239.3b _»

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


Apple (5A) 917.9 (10) 58.9 246.5
Apple (5B) 1449 (3) 680.3b 147.5 _J> 696.1
Swordfish (4) 2377 (19) 1114.7s 230.1 -J> 1138.2
NBS Tuna (2) 3061 (19) 597.7 ' 397.8 337.6 686.4
NBS Orchard Leaves (1) 9265 (16) 184.7 569.0 600.0 826.9 847.3
IAEA oyster (3/13) 11447 (18) 1298.1 c 992.8 912.3 1586.6
Flounder (6) 13149 (15) 4265.2* 736.2 _& 4328.2

" Unconnected for moisture c o n t e n t ; n u m b e r of laboratories contributing t o t h e average is indicated in paren-


theses.
b
SL, SAand sa cannot be calculated; value between t h e s t a n d SA columns is V S L 2 + SA 2 .
c
Variance is negative; standard deviation is imaginary.

Table 5. Statistical evaluation of collaborative results for selenium (ng/g)

Between-lab. Between-dig. Within-dig. Precision Reprod.


Sample Av. concn std dev., std dev., std dev., std dev., std dev.,
(code No.) found" SL SA SD SH Si
Apple (5A) 31.0(10) 79.5b 30.9 —* 85.3
NBS Spinach (9) 42.9(12) 26.4 —" 34.4 29.4 39.5
NBS Orchard Leaves (1) 78.0(19) 53.3 40.3 35.7 53.8 75.7
Skim milk powder (11) 82.9(14) 23.2b 12.0 —h 26.1
Kale (7/14) 131.5(16) 40.9 13.1 23.5 26.9 49.0
Apple (5B) 160.3(4) 104.5b 77.9 —* 130.4
IAEA flour (10) 213.1(18) 83.0* 64.3 —b 104.9
Flour (12) 686.9(18) 68.5b 71.5 —* 99.0
NBS Bovine Liver (8) 774.5(19) 348.5 107.3 125.8 165.4 385.7
Flounder (6) 1319 (18) 232.46 156.0 —b 279.9
IAEA oyster (3/13) 1447 (19) 596.5 179.6 175.2 250.9 647.1
Swordfish (4) 2538 (17) 325.7* 189.4 —b 376.8
NBS Tuna (2) 3625 (19) 689.9 —c 293.3 276.9 743.4
b c
"• - See Table 4.

Table 6. Coefficients of variation for arsenic Table 7. Coefficients of variation for selenium

Av. concn. C o e f f . o f v a r ._
Av. concn Coeff. of var.,
Sample found, Sample found,
(code No.) ng/g" (code No.) ng/g" SR sx
Skim milk powder Apple (5A) 31.0 275.2
(11) 0.12 22500 NBS Spinach (9) 92.1
42.9 68.5
IAEA flour (10) 31.2 189.4 NBS Orchard
Flour (12) 31.4 209.6 Leaves (1) 69.0 97.1
78.0
NBS Bovine Liver Skim milk powder
(8) 43.3 62.4 86.4
(11) 82.9 31.5
Kale (7/14) 93.1 27.9 39.7
Kale (7/14) 131.5 20.5 37.3
NBS Spinach (9) 114.0 17.8 43.7
81.3
917.9 26.9
Apple (5B) 160.3
Apple (5A) 49.2
1449 IAEA flour (10) 213.
Apple (5B) 48.0 14.4
2377 Flour (12) 686.
Swordfish (4) 47.9
3061 NBS Bovine Liver
NBS Tuna (2) 11.0 22.4
(8) 774. 21.4 49.8
NBS Orchard
Flounder (6) 1319 21.2
Leaves (1) 9265 8.9 9.1 IAEA oyster (3/13) 1447 17.3 44.7
IAEA oyster (3/13) 11447 8.0 13.9 Swordfish (4) 2538 14.8
Flounder (6) 13149 32.9 NBS Tuna (2) 3625 7.6 20.5
" Uncorrected for moisture content.
b
SH cannot be calculated for these samples. «•' See Table 6.
Table 8. Reference analyses (ng/g) of collaborative samples for arsenic9
Method and laboratory

NAA with cheim. Light absorption Light absorption Light absorption Graphite
b Other values
sepn— IDSSMS^- photometry with photometry with phototnetrv with furnace AAS&
(n) (n) ^ (n) AgDDC^ AgDDC^ (n)
Sample
code

10900 (1) 10000+2000 1 : 1060O+800(ll)i; 11000+1000(12)-;


9000(13)i; 13000(13)E

2 1100+100 (2) 1500+300 (S) 4200+500 (S) 1200 3200 (1) 1740 4600+300(11)-; 3600+200(12)°; 3300+200(12)-
4600+300(1 §
3)
3/13 6200+300 (2) 17000+900 (3) 8700 1290±200 (2) 10700 >
4 700+150 (2) 1900+300 (i) 7600 2700+100 (2) 1640 3360+4O3- o
6 5900+300 (2) 6500+1000 (2) 19400+1000^ (6) 7700 11800+200 (2) 9370 157OO+25O3-

>
a
7/14 100+20 (2) <50 260 127+29(16)-; 141 (range 110-220) (17,18)-

5^-; 49+6(12)- o
9 90+20 (2) <iooo 150+50^; 198+10^

10 10+5 (2) <iooo 50 24+8(19)-


p
11 <5 (2) <50

12 <3 (2) <50

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020

» «•-»
° All concentrations, initially reported on an as-received basis, have been corrected to a dry sample basis; all "other values" are also presumably on a dry basis.
Mean concentrations ± standard error of n analyses/determinations are listed. Values chosen for the calculation of selected reference levels are underscored.
6
P. Schramel; neutron activation analysis with radiochemical separation after irradiation (6).
" J. A. Carter and D. L. Donohue; low temperature ashing, isotope dilution spark source mass spectrometry (7), using 11000 ng As/g of NBS Orchard Leaves SRM 1571 8-
as reference value.
d
M. Stoeppler and W. Matthes; digestion with HNO3-HCIO4-H2SO4 as described for this collaborative study, and photometry using AgDDC (8, 9).
' E. Sandi and M. T. Lo; ashing at 480°C with Mg(NO3>2, and photometry using AgDDC according to modified AOAC method (10).
1
K. H. Tarn; ashing at 500°C with Mg(NO 3 ) 2 and MgO, and photometry using AgDDC.
' R. B. Flemming and H. C. Freeman; digestion with HNO3, determination using Perkin-Elmer HGA-2000 graphite furnace.
* Concentration (ng/g) ± uncertainty as defined in footnotes; numbers in parentheses refer to references at end of paper.
' NBS certified value ± twice standard deviation or entire range of observed results. o
' Mean ± standard deviation of >4 determinations using neutron activation analysis with chemical separation after irradiation. 3
* Mean ± standard deviation based on analysis of 3 replicate 1 g samples using hydride AAS method. a
1 a
Dry ashing/hydride AAS method following procedures of Dalton and Malanoski (14, 15); no uncertainty stated.
m
Value obtained by another laboratory using an unspecified AAS method.
" N B S informational value, mean ± standard deviation of analyses of 8 samples, using neutron activation analysis with chemical separation.
° FDA instrumental neutron activation value; uncertainty includes counting statistics.
"Sample still under analysis in IAEA intercomparison program; IAEA value is not yet available. n
* H. J. Miller; hydride AAS; mean ± standard error. a
' Sample 6 was sent to this laboratory under 2 different codes. o
* Best mean ± standard deviation; 16 activation analyses by 3 different laboratories.
o
1
Best value and range of laboratory means; 27 activation analyses by 7 laboratories.
" Uncertified NBS value.
" IAEA value ± t X standard error for 95% confidence limits based on 10 analyses by 3 laboratories using neutron activation analysis and light absorption photo-
metry. Value reported by IAEA (19) corrected to dry basis.
O
00
H

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


to
00

Table 9. Reference analyses <ng/g) of collaborative samples for selenium"


Method and laboratory

NAA with chem. INAA~ (n) INAA- (n) t u v w


sepnS- (n) Fluorometry— Fluorometry Fluorometry" Fluorometry— Other values
Sample (n) (n) (n) (n)
code

95+3 (3) 79+2 (4) 80+10^ 80+10(ll)J-; 55+9(12)-; 118+17-;


Tr<5)^ ~
2 23900+2100 (2) 3850+120 (2) 4430+210 (3) 4180+50 (S) 3710+60 (4) 3710+130 (2) 3300+300(11)-; 3600+100(12)-; 3720+1705-

3/13 1200+210 (2) 1830+150 (2) 2140+210 (3) 2230+40 (5) 2010+50 (6) 1880+50 (2) 2220+30 (6)

4 1100+200 (2) 2900+60 (2) 3270+200 (3) 2950+40 (S) 2830+70 (6) 2360+170 (4) 2900+50 (2) 2910+45(5)i

5A 5+1 (4) 66+16^; 23+2^

5B 6+3 (4) 1+1 (2) 203+93.

6 660+150 (2) 1430+70 (2) 1630+200 (3) H40+40 (5) 1400+30 tb) 1420+180 (4) 1470+15 (4) 1450+50^; 1580+203- W
7/14 140+15 12) <2OO+2OO (2) <200 (3) 125+4 (S) 135+2 (6) 122+10 (3) 126+2 (6) 200+52-; 148+14(16)—; 15O?(17)—;
121?140 (range 20-150) (18)-££- I
1260+40 (S) 1220+5 (4) UOQ+100^-; 1090+50(11)^-; 980+10(12)-;
+^0^; 1210+30(5)2.

9 <100 (2) <30O+300 (2) <200 (3) 45+10 (4) 38+1 (6)

10 370+60 (2) 260+20 (2) 270+40 (3) 295+10 (S) 280+4 (6) 235+20 (4) 286+5 (4) 187+34-
p
11 <30 (2) <200+200 (2) <100 (3) 97+2 IS) 100+2 IB) 85+3 (2) 99+1 f4J 112+6(5)^

12 630+110 (2) 750+30 (2) 790+40 (3) 860+10 (4) 805+15 (6) 830+10 (2) 850+15 14) 804+12(5)^

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020

^ i
« .r

°-« See Table 8.


r
J. F. Emery; instrumental neutron activation analysis. a
* J. T. Tanner and M. H. Friedman; instrumental neutron activation analysis (20, 21).
' R. A. Cloutier; fluorometry using 2.3-diaminonaphthalene and AOAC method 25.117-25.120. The same reagent was used by all other analysts using fluorometry.
* J. H. Watkinson and M.T. Wingerden; digestion with HNO3-HCIO4, determination using semiautomated adaptation of fluorometric method of Watkinson (22).
* A. Sosner and R. Cohen; fluorometry, AOAC method 25.117-25.120 with omission of H2O2.
" J. Jago; digestion with HNO3-HCIO4, fluorometry according to method of N. L. Wilson (23), essentially the method of Watkinson (22). f
* M. Ihnat; hydride AAS; mean ± standard error.
'Mean ± standard error; based on 14 (Sam pie 1), 16 (Samples 4, 8, and 12), and 18 (Sample 11) analyses by as many laboratories during collaborative study of f luoro- r°
metric method, 25.117-25.120 (5). a
' Sample still under analysis in IAEA intercomparison program; IAEA value is not yet available; the listed value was determined by M. Ihnat (seex). o
"" Mean ± standard deviation based on 20 analyses by 4 laboratories using fluorometry and activation analysis. S
66
Best value. Agreement between results from activation analysis (ranges of means from 4 laboratories, 17-620 ng/g) and fluorometry (individual means from 2 labo- 3
ratories, 140 and 140 ng/g) is poor. H
" Best mean, most probable value, and range, respectively. Bowen writes " . . . for kale the analyses show a bimodal distribution with two 'most probable' values
around 40 and 140 ng/g. Further work is needed to establish which of these is correct but it is suspected that the volatility of the element may have given rise to the CO

lower values."
dd
NBS certified value with 95% confidence limits estimate of uncertainty.
O
d

GO

Z
u

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


1430 JOURNAL OF THE AOAC (Vol. 60, N o . 6, 1977)

Table 10. Accuracy of the hydride AAS method 11 as well as those in Tables 8 and 9 have been
for arsenic" corrected for sample moisture contents. Com-
Selected ref., Av. found, parisons for arsenic (Table 10) could only be
Sample ng/g 6 ng/g" made for 6 of the 13 samples for the reason
IAEA flour 24 ±8 (3)"1 35 ±17 (14)
given above. Agreement was excellent between
NBS Bovine Liver 55 (NBS) 6
46 ±8 (17) the hydride and NBS values for Orchard Leaves.
Bowen's kale 141 ±15 (7) 101 ±9 (17) Remarkedly good agreement was obtained for
NBS Spinach 150±5O(NBSy 120 ±13 (16)
NBS Tuna 3900 ±300 (4) 3160 ±145 (19)
IAEA flour, NBS Bovine Liver, kale, and NBS
NBS Orchard Leaves 10000±2000 (NBS)/ 9940 ±145 (16) Spinach in spite of the low arsenic levels present

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


in these examples; agreement was poorer for
" All results are on a dry sample basis and are re- NBS Tuna. For 5 samples (except IAEA flour),
ported as mean ± standard error (number of labora-
tories), unless otherwise noted. however, hydride AAS values were consistently
b
Reference levels were calculated by averaging the lower.
selected data of Table 8 giving identical weights to the
mean results from each laboratory. Several samples For selenium (Table 11), some sort of refer-
are not listed because agreement among reference ence numbers were available for all 13 samples.
laboratories was poor, making it impossible to select
data. For 8 samples, NBS Spinach, NBS Orchard
c
Standard error, sx is defined in text. Leaves, skim milk powder, kale, IAEA flour,
d
Uncertainty reported by IAEA is ± t X standard flour, flounder, and NBS Tuna, the accuracy of
error for 95% confidence level.
* Tentative NBS level. the hydride AAS method was very good. Hy-
1
NBS certified concentration including twice the dride AAS results for NBS Bovine Liver, IAEA
standard deviation or the entire range of observed oyster, and swordfish were, respectively, 25, 24,
results.
and 11% lower than selected reference levels.
Firm comparisons between data were not made
and Spinach, are certified for arsenic) made for the 2 apple samples because only limited
these analyses, together with other reported val- data were available from reference laboratories,
ues, invaluable (at least for selenium) in arriv- and collaborators analyzed spiked samples,
ing at selected reference estimates of levels in a whereas reference laboratories analyzed unspiked
number of samples. samples. The 3 samples exhibiting low recoveries
Only levels generated by methods other than are either meat or fish, suggesting a systematic
hydride AAS formed the basis for selection; sample matrix effect either in hydride evolution
non-collaborative results obtained by hydride or sample treatment. Curiously, tuna, a sample
generation AAS are included in Tables 8 and 9 also of marine origin as oyster and swordfish,
only for information. Seemingly acceptable val- showed good agreement of determined selenium
ues were chosen by inspection and pooled to levels. Average selenium levels found in the
give mean selected levels shown in Tables 10 and 1973 collaborative study of a fluorometric method
11. For NBS Orchard Leaves, Bovine Liver, and for selenium (5) are also included in Table 11
Spinach, the NBS certified levels were used for information. Comparison of these data with
where available. Large disagreement among ref- either NBS values or other appropriately selected
erence results for arsenic made selection difficult reference levels (excluding the same fluorometric
or impossible with the consequence that very data) demonstrates the good accuracy of fluoro-
few reference results are available for this ele- metric analysis. These tests for accuracy indi-
ment ; its determination seems difficult. For sele- cate that for both elements, accuracy of the
nium, on the other hand, between-laboratory hydride AAS method is quite acceptable for a
and between-method agreement is excellent in number of different (but not all) sample types,
comparison, and many data were available for in spite of the presence of a large variance asso-
estimating selenium levels. ciated with systematic error.
Comparisons of arsenic and selenium concen- In arriving at the foregoing observations of
trations obtained with the hydride AAS method, method performance, it must be kept in mind
with selected reference levels, are presented in that the data were screened and a number of
Tables 10 and 11. Concentrations listed here do outlying observations have been rejected prior
not agree with the corresponding ones in Tables to final calculations. Whereas the rejection of
4-7 because the concentrations in Tables 10 and some results seemed obvious, the perpetual ques-
IHNAT & MILLER: HYDRIDE AAS METHOD FOR AS AND SE IN FOOD 1431

Table 11. Accuracy of the hydride AAS method for selenium0

Fluorometric
Selected ref., Av. found, method (5),
Sample ng/g 6 ng/g c ng/g

Mclntosh apple 5 ± 1 ay 32 ±27 doy


Delicious apple 4 ±4 (2)" 170 ±63
NBS Spinach 42 ±4 (2) 45 ±9 (12)
w
NBS Orchard Leaves 80 ±10 (NBS)/ 84±15 (19) 92 ±5 (14)
Skim milk powder 109 ±5 (22) 87 ±7 (14) 112 ±6 (18)
[95±3 (4)]»

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


Bowen's kale 135 ±3 (8) 143 ±12 (16)
IAEA flour 271 ±9 (6) 239 ±25 (18)
Flour 807 ±11 (22) 760 ±22 (18) 804±12 (16)
[814 ±17 (6)]'
NBS Bovine Liver 1100 ±100 (NBSy 826 ±89 (19) 1200 ±30 (16)
Flounder 1470 ±34 (6) 1350 ±62 (18)
IAEA oyster 2050 ±160 (6) 1550 ±150 (19)
Swordfish 2900 ±46 (22) 2590 ±87 (17) 2910 ±45 (16)
[2870 ±120 (6)]»
NBS Tuna 3860 ±160 (6) 3740 ±170 (19)

• All results are on a dry sample basis and are reported as mean ± standard error (number of laboratories),
unless otherwise noted.
b
Reference levels were calculated by averaging the selected data of Table 9 giving identical weights to the mean
(or only) results from each laboratory. Two low values of 17and 46 ng/g and one high value of 620 ng/g determined
by activation analysis for kale (17) were rejected.
e
Standard error s i is defined in text.
d
Insufficient data.
' These samples were fortified by collaborators using solution from ampoule D whereas reference analyses were
done on non-fortified samples.
1
NBS certified concentration including 95% confidence limits or entire range of observed results.
' Reference levels calculated by omitting data from collaborative study of fluorometric method (5).

tion arises as to whether rejected values were precision. Recommendations are contained in the
bona fide members of the data population or following report.
aberrant data. Some results such as the 0 ng
As/g reported by Collaborator 49 for several Acknowledgments
samples are likely to be aberrant values but The Associate Referees are deeply grateful to
could conceivably arise from, and flag a weak the following for their efforts in contributing to
point in the method; the assessment of method the study as collaborators and reference labo-
performance characteristics should be tempered ratories :
by this behavior and an attempt should be made R. C. Abel, Food and Drug Administration,
to locate the cause of the problem. The method Baltimore, MD
cannot be recommended if some explanation F. A. J. Armstrong and A. Lutz, Environment
cannot be found for the large numbers of out- Canada, Freshwater Institute, Winnipeg, Mani-
lying observations. Whether in fact the method toba, Canada
or laboratory-related techniques are responsible R. Arreola, R. R. Cortez, D. F. Fischer, and
for the observed performance is a question to be W. R. Holtz, Joint Water Pollution Control
tackled in the detailed treatment of data pre- Plant, Water Quality Laboratory, Carson, CA
sented in the following reports (3, 4). E. E. Bailey, and J. P. Minyard, Jr, Missis-
In summary, this study has shown that under sippi State Chemical Laboratory, Mississippi
conditions of a specified sample treatment pro- State, MS
cedure and essentially free choice of hydride Z. Baraniak, B. J. Houlahan, and D. Kova-
generation equipment and technique, the per- chic, Agriculture Canada, Plant Products Divi-
formance of an acid-digestion NaBH4-facilitated sion, Ottawa, Ontario, Canada
hydride generation AAS method for estimating R. A. Buckley and G. B. Jones, CSIRO, Divi-
arsenic and selenium in foods is fairly good with sion of Human Nutrition, Adelaide, Australia
respect to accuracy for many, but not all, food S. G. Capar and J. W. Jones, Food and Drug
samples tested and unacceptable with respect to Administration, Washington, DC
1432 JOURNAL or THE AOAC (Vol. 60, No. 6,1977)

J. A. Carter, D. L. Donohue, and J. F. Emery, and Welfare Canada, Food Directorate, Ottawa,
Oak Ridge National Laboratory, Analytical Ontario, Canada
Chemistry Division, Oak Ridge, TN B. Loescher, Ontario Ministry of the Environ-
C. Y. Chan and P. N. Vijan, Ontario Min- ment, Laboratory Services Branch, Rexdale, On-
istry of the Environment, Air Quality Labora- tario, Canada
tory, Toronto, Ontario, Canada W. Matthes and M. Stoeppler, Nuclear Re-
G. D. Christian and E. J. Knudson, Depart- search Centre, Institute of Chemistry, 4 Jue-
ment of Chemistry, S. Olsen and W. R. Schell, lich, West Germany
Laboratory of Radiation Ecology, University of R. W. Marts, Food and Drug Administration,

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


Washington, Seattle, WA Kansas City, MO
0. E. Clinton, J. H. Watkinson, and M. T. C. M. McGregor and J. Parker-Sutton, Min-
Wingerden, New Zealand Ministry of Agricul- istry of Agriculture, Fisheries, and Food, Derby,
ture and Fisheries, Ruakura Agricultural Re- UK
search Centre, Hamilton, New Zealand J. C. M. Mol-Krijnen, B. E. Murphy, and
R. A. Cloutier, Agriculture Canada, Chemistry S. R. B. Solly, New Zealand Ministry of Agri-
and Biology Research Institute, Ottawa, On- culture and Fisheries, Wallaceville Animal Re-
tario, Canada search Centre, Upper Hutt, New Zealand
R. Cohen and A. Sosner, Folkman and Dr. S. M. Muir and F. J. Schmidt, Illinois Envi-
Koffler Ltd., Tel Aviv, Israel ronmental Protection Agency, Champaign, IL
T. T. Collins, M. A. Nevelle, T. Surtes, and W. Pfannhauser and H. Woidich, Forschungs-
J. R. Tuschall, Jr, CDM/Limnetics, Milwaukee, institut der Ernahrungswirtschaft, Vienna,
WI Austria
E. J. Dixon, Laboratory of the Government R. V. D. Robert, National Institute for Metal-
Chemist, Department of Industry, London, UK lurgy, Johannesburg, South Africa
R. J. Emerick and 0. E. Olson, South Dakota P. Schramel, Gesellschaft fur Strahlen-und
State University, Department of Chemistry, Umweltforschung mbH., Neuherberg/Munchen,
Brookings, SD West Germany
F. E. Estiandan, M. Shimabukuro, and N. A. M. Sulek, National Canners Association,
Uichanco, Carnation Research Laboratories, Van Washington, DC
Nuys, CA J. R. Williams, Food and Drug Administra-
F. J. Fernandez, The Perkin-Elmer Corp., tion, San Francisco, CA
Norwalk, CT R. J. Westerby, Agriculture Canada, Chem-
R. B. Flemming and H. C. Freeman, Environ- istry and Biology Research Institute, Ottawa,
ment Canada, Fisheries and Marine Service, Ontario, Canada
Halifax, Nova Scotia, Canada- The authors are indebted to the following who
M. H. Friedman and J. T. Tanner, Food and supplied on short notice most of the samples
Drug Administration, Washington, DC used in the study: H. J. M. Bowen, Department
L. R. Hageman, State of Montana Depart- of Chemistry, The University, Reading, UK
ment of Agriculture, Montana State University, (kale); International Atomic Energy Agency,
Bozeman, MT International Laboratory of Marine Radioac-
M. B. Hocking and M. Ko, University of Vic- tivity, Monaco-ville, Monaco (oyster), Analyti-
toria, Department of Chemistry, Victoria, Brit- cal Quality Control Services, Vienna, Austria
ish Columbia, Canada (flour); U.S. Food and Drug Administration,
Y. Inoue, M. Munemori, and T. Nakahara, Baltimore, MD (NBS Bovine Liver and Orchard
University of Osaka Prefecture, College of Engi- Leaves); J. T. Tanner, U.S. Food and Drug Ad-
neering, Mozu-umemachi, Saki 591, Japan ministration, Washington, DC (NBS Spinach
R. A. Isaac and W. C. Johnson, University of and Tuna); and J. Graham, Environment Can-
Georgia, Soil Testing and Plant Analysis Labo- ada, Ottawa and K. Spencer, Environment Can-
ratory, Athens, GA ada, St. John's, Newfoundland (flounder). The
J. Jago, Government Chemical Laboratories, technical assistance of R. J. Westerby, Agricul-
Perth, Australia ture Canada, Ottawa, R. C. Abel, U.S. Food and
M. T. Lo, E. Sandi, and K. H. Tam, Health Drug Administration, Baltimore, MD, and J.
IHNAT & MILLER: HYDRIDE AAS METHOD FOR AS AND SE IN FOOD 1433

Jones, Agriculture Canada, Food Research Insti- (12) Fiorino, J. A., Jones, J. W., & Capar, S. G.
tute, Ottawa (for freeze-drying several fish sam- (1976) Anal. Chem. 48, 120-125
ples) is appreciated. Moisture determinations (13) Zook, E. G., Powell, J. J., Hackley, B. M.,
were performed by Analytical Chemistry Serv- Emerson, J. A., Brooker, J. R., & Knobl,
ices, Chemistry and Biology Research Institute, G. M., Jr (1976) J. Agric. Food Chem. 24, 47-
53
Agriculture Canada. The assistance of B.
(14) Dalton, E. F., & Malanoski, A. J. (1969)
Thompson, Statistical Research Service, Agricul- JAOAC 52, 1035-1038
ture Canada, with the design of the study and (15) Dalton, E. F., & Malanoski, A. J. (1971) At.
the statistical computations is gratefully ac- Absorp. Newsl. 10, 92-93

Downloaded from https://academic.oup.com/jaoac/article/60/6/1414/5710230 by guest on 04 November 2020


knowledged. (16) Bowen, H. J. M. (1967) Analyst 92, 124-131
(17) Bowen, H. J. M. (1969) Adv. Act. Anal. 1,
REFERENCES 101-113
(1) Official Methods of Analysis (1975) 12th Ed., (18) Bowen, H. J. M. (1974) J. Radioanal. Chem.
AOAC, Washington, DC 19, 215-226
(2) Ihnat, M., & Miller, H. J. (1977) 60, 813-825 (19) Gorski, L., Heinonen, J., & Suschny, O.
(3) Ihnat, M., & Thompson, B. (1977), in prepara- (1974) "Final Report on the Intercomparison
tion of Trace Multielement Analysis in Dried Ani-
(4) Thompson, B., & Ihnat, M. (1977), in prepara- mal Whole Blood (A-2), Calcinated Animal
tion Bone (A-3/1), Powdered Milk (AS), Wheat
(5) Ihnat, M. (1974) JAOAC 57, 373-378 Flour (V-2/1), and Dried Potatoes (V-4),"
(6) Samsahl, K. (1967) Anal. Chem. 39, 1480-1483 International Atomic Energy Agency, Vienna,
(7) Carter, J. A., Donohue, D. L., Franklin, J. C , Austria, No. IAEA/RL/25
& Stelzner, R. W. (1975) 9th Annual Con- (20) Friedman, M. H., & Tanner, J. T. (1975)
ference on Trace Substances in Environ- Trans. Am. Nucl. Soc. 21, 97-98
mental Health, University of Missouri, Co- (21) Friedman, M. H., & Tanner, J. T. (1975) / .
lumbia, MO, June 9-12, pp. 303-310 Radioanal. Chem. 25, 269-273
(8) Woidich, H., & Pfannhauser, W. (1975) Z. (22) Watkinson, J. H. (1966) Anal. Chem. 38, 92-
Lebensm. Unters. Forsch. 159, 323-327 97
(9) Dahl, R. (1973) Int. Ber. der KFA Juelich (23) Wilson, N. L. (1969) "The Determination of
(unpublished). Selenium in Agricultural Materials," Report
(10) Official Methods of Analysis (1975) 12th Ed., of Investigations No. 6, Government Chemi-
AOAC, Washington, DC, sec. 25.012 cal Laboratories, Perth, Western Australia
(11) Orvini, E., Gillis, T. E., & LaFleur, P. D.
(1974) Anal. Chem. 46, 1294-1297 Received December 15, 1976.

You might also like